Information recommendation methods, apparatus, computing devices, and computer-readable storage media

Abstract

This disclosure provides an information recommendation method. This method dynamically determines whether repeatedly recommended information should continue to be displayed by setting an estimated attractiveness index for the candidate information when it is repeatedly recommended. Furthermore, when adjusting candidate information, this method comprehensively considers information freshness indicators and indicators reflecting platform revenue, thereby minimizing the adverse impact of freshness strategies on revenue. Thus, the information recommendation method according to this disclosure can establish a mutually satisfactory balance between user experience and business revenue. In addition, this disclosure also relates to information recommendation devices, computing devices, computer-readable storage media, and computer program products employing this method.

Description

Technical Field

[0001] This disclosure relates to the field of computer technology, and more specifically, to an information recommendation method and apparatus, a computing device, and a computer-readable storage medium. Background Technology

[0002] With the popularization of the internet and the development of network technology, more and more businesses are conducted online. During these transactions, the business side typically recommends information to users. However, in situations where information recommendations are frequent—for example, when the time interval between consecutive user requests for information recommendations is short—multiple identical pieces of information are often recommended and displayed within a short period.

[0003] On the one hand, when users are genuinely uninterested in the information, excessive repetition of the same information can significantly disrupt their experience, thus degrading user engagement and wasting resources for the business. On the other hand, when users are interested in the information and have already clicked or converted, subsequent repetition can also be disruptive. Moreover, allocating subsequent recommendation / display opportunities to other information that may lead to clicks or conversions could potentially yield greater overall benefits for the business.

[0004] In related technologies, to reduce user aversion to the repeated display of the same information, business operations typically employ a freshness strategy to control the repeated recommendation of identical information. While the freshness strategy can significantly reduce the repeated recommendation / display of the same information, it still has some shortcomings. For example, it may affect the revenue of the business, potentially have a negative impact on the information provider, and fail to enhance the impact on user perception, among other things. Summary of the Invention

[0005] In view of the above, this disclosure provides an information recommendation method and apparatus, which are intended to overcome some or all of the defects mentioned above, as well as other possible defects.

[0006] According to one aspect of this disclosure, an information recommendation method is provided. The method includes the following steps: in response to an information recommendation request received from a terminal device, generating a candidate information queue, the candidate information queue including first candidate information at a first sorting position; when the first candidate information relates to the same recommended object as recommendation information recommended to the terminal device within a predetermined time period prior to the current time, determining an estimated attractiveness of the first candidate information, the estimated attractiveness being the probability that a user of the terminal device will perform a desired action based on the first candidate information; and recommending the first candidate information when the estimated attractiveness is greater than or equal to a predetermined attractiveness threshold.

[0007] According to some exemplary embodiments, in response to an information recommendation request received from a terminal device, a candidate information queue is generated, the candidate information queue including first candidate information at a first sorting position, including: sorting each candidate information based on expected revenue per thousand exposures to generate the candidate information queue, wherein the first candidate information has the largest expected revenue per thousand exposures.

[0008] According to some exemplary embodiments, in response to an information recommendation request received from a terminal device, a candidate information queue is generated, the candidate information queue including first candidate information at a first sorting position, including: sorting the candidate information based on traffic value to generate the candidate information queue, wherein the first candidate information at the first sorting position has the largest traffic value.

[0009] According to some exemplary embodiments, when the first candidate information involves the same recommended object as the recommendation information recommended to the terminal device during a predetermined time period before the current time, the estimated attractiveness of the first candidate information is determined. The estimated attractiveness is the probability that the user of the terminal device will perform a desired action based on the first candidate information. This includes: determining the estimated click-through rate of the first candidate information and using the estimated click-through rate as the estimated attractiveness.

[0010] According to some exemplary embodiments, when the first candidate information involves the same recommended object as the recommendation information recommended to the terminal device during a predetermined period before the current time, the estimated attractiveness of the first candidate information is determined, wherein the estimated attractiveness is the probability that the user of the terminal device will perform a desired action based on the first candidate information, including: determining the estimated conversion rate of the first candidate information and using the estimated conversion rate as the estimated attractiveness.

[0011] According to some exemplary embodiments, the information recommendation method further includes: when the estimated attractiveness is less than the predetermined attractiveness threshold, determining the information freshness of each candidate information in the candidate information queue, wherein the information freshness is the overall similarity between each candidate information and at least one piece of information that has been displayed by the terminal device before the current moment; reordering each candidate information based on the information freshness of each candidate information in the candidate information queue and the corresponding expected revenue per thousand exposures to obtain a first reordered candidate information in the candidate information queue; and recommending the first reordered candidate information.

[0012] According to some exemplary embodiments, when the estimated attractiveness is less than the predetermined attractiveness threshold, the information freshness of each candidate information in the candidate information queue is determined. This information freshness is the overall similarity between each candidate information and at least one piece of information already displayed by the terminal device before the current moment, including: obtaining the previous... nThe information already displayed in this session includes... n The integer is greater than 0; based on the information already shown, construct... n × n Similarity matrix S Among them, the matrix elements of the similarity matrix S ,in α ij Indicates the first i The information already displayed is related to the first item. j The similarity of the displayed information is 0 < α ij ≤1, and i and j All are integers, 0 < i ≤ n ,0< j ≤ n , and s ij =s ji ; Traverse each candidate information in the candidate information queue and construct an augmented similarity matrix corresponding to each candidate information. ,in, S This is the similarity matrix. V It is the similarity vector of the candidate information relative to the already displayed information. V =[ v 1 , v 2 , ... , v n ],in v 1 , v 2 , ... , v n This candidate information is respectively with n The similarity of the displayed information, and v 1 , v 2 , ... , v n Each element in the matrix is ​​greater than 0 and less than or equal to 1, and T represents the transpose of the matrix; and the augmented similarity matrix of each candidate information in the candidate information queue is calculated by performing a determinant calculation, and the values ​​obtained by the determinant calculation are used as the information freshness of each candidate information.

[0013] According to some exemplary embodiments, the similarity between two pieces of information is calculated as equal to ,in, m An integer greater than or equal to 0, representing the direct, non-repeating distance between the two pieces of information.

[0014] According to some exemplary embodiments, in this similarity matrix, when i ≠ j When, 0 < α ij <1.

[0015] According to some exemplary embodiments, based on the information freshness of each candidate information in the candidate information queue and the corresponding expected revenue per thousand impressions, the candidate information is reordered, including: multiplying the information freshness of each candidate information in the candidate information queue by the corresponding expected revenue per thousand impressions or traffic value to obtain the corresponding recommendation index; and taking the candidate information with the highest recommendation index among the candidate information in the candidate information queue as the first reordered candidate information.

[0016] According to another aspect of this disclosure, an information recommendation apparatus is provided. The apparatus includes: a candidate information queue generation module configured to: generate a candidate information queue in response to an information recommendation request received from a terminal device, wherein the candidate information queue includes first candidate information at a first sorting position; an estimated attractiveness determination module configured to: determine an estimated attractiveness of the first candidate information when it relates to the same recommended object as recommendation information recommended to the terminal device within a predetermined time period prior to the current time, the estimated attractiveness being the probability that a user of the terminal device will perform a desired action based on the first candidate information; and a recommendation module configured to: recommend the first candidate information when the estimated attractiveness is greater than or equal to a predetermined attractiveness threshold.

[0017] According to some exemplary embodiments, the information recommendation device further includes: an information freshness determination module, configured to: determine the information freshness of each candidate information in the candidate information queue when the estimated attractiveness is less than the attractiveness threshold, wherein the information freshness is the overall similarity between each candidate information and at least one piece of information displayed by the terminal device before the current moment; and a reordering module, configured to: reorder each candidate information based on the information freshness of each candidate information in the candidate information queue and the corresponding expected revenue per thousand exposures, to obtain a first reordered candidate information in the candidate information queue; wherein the recommendation module is further configured to: recommend the first reordered candidate information.

[0018] According to some exemplary embodiments, the information freshness determination module includes: a displayed information acquisition submodule, configured to: acquire the previous information of the terminal device from the current moment. n The information already displayed in this session includes... n The integer is greater than 0; the similarity matrix generation submodule is configured to: construct based on the displayed information. n × n Similarity matrixS The similarity matrix S Each matrix element ,in α ij Indicates the first i The information already displayed is related to the first item. j The similarity of the displayed information is 0 < α ij ≤1, and i and j All are integers, 0 < i ≤ n ,0< j ≤ n , and s ij =s ji The augmented similarity matrix generation submodule is configured to: traverse each candidate piece of information in the candidate information queue and construct an augmented similarity matrix corresponding to each candidate piece of information. ,in, S This is the similarity matrix. V It is the similarity vector of the candidate information relative to the already displayed information. V =[ v 1 , v 2 , ... , v n ],in v 1 , v 2 , ... , v n This candidate information is respectively with n The similarity of the displayed information, and v 1 , v 2 , ... , v n Each element in the matrix is ​​greater than 0 and less than or equal to 1, where T represents the transpose of the matrix; and the information freshness generation submodule is configured to: perform determinant calculation on the augmented similarity matrix of each candidate information in the candidate information queue, and use the values ​​obtained from the determinant calculation as the information freshness of each candidate information.

[0019] According to some exemplary embodiments, the similarity between two pieces of information is calculated as equal to ,in, m An integer greater than or equal to 0, representing the direct, non-repeating distance between the two pieces of information.

[0020] According to some exemplary embodiments, in the similarity matrix and the augmented similarity matrix, when i ≠ j When, 0 < α ij <1.

[0021] According to some exemplary embodiments, the reordering module includes: a recommendation index generation submodule, configured to: multiply the information freshness of each candidate information in the candidate information queue by its corresponding expected revenue per thousand impressions or traffic value to obtain the corresponding recommendation index; and a first reordering candidate information determination submodule, configured to: take the candidate information with the largest recommendation index among the candidate information in the candidate information queue as the first reordering candidate information.

[0022] According to another aspect of this disclosure, a computing device is provided. The computing device includes a processor and a memory configured to store computer-executable instructions configured to cause the processor to perform the information recommendation method described above when executed on the processor.

[0023] According to another aspect of this disclosure, a computer-readable storage medium is provided. The computer-readable storage medium is configured to store computer-executable instructions configured to cause the processor to perform the information recommendation method described above when executed on a processor.

[0024] According to another aspect of this disclosure, a computer program product is provided, including computer-executable instructions configured to cause the processor to perform the information recommendation method as described above when executed on a processor.

[0025] The beneficial effects of the scheme disclosed herein include at least the following aspects:

[0026] Firstly, for information recommendation platforms, compared to existing freshness strategies based on frequency control, this solution dynamically determines whether the same information should continue to be displayed by setting an estimated attractiveness index for information. This only intervenes in low-value duplicate information while allowing valuable information to be recommended repeatedly, thus having a smaller impact on platform revenue. It achieves the goal of reducing the overall rate of recommendation duplication while increasing the click-through rate and traffic value.

[0027] Secondly, for users, this solution reduces the repeated recommendations of the same information, thereby reducing invalid displays on user devices, improving user experience, and in the long run, benefiting user activity and retention.

[0028] Finally, for information providers, this solution optimizes the high costs caused by invalid displays due to duplicate recommendations, thus reducing the information provider's campaign costs and improving campaign efficiency. Attached Figure Description

[0029] The specific embodiments of this disclosure will now be described in detail with reference to the accompanying drawings, so as to enable a fuller understanding of the further details, features and advantages of this disclosure, in which:

[0030] Figure 1 This illustration schematically depicts a general application scenario of a business system according to some exemplary embodiments of this disclosure;

[0031] Figure 2 The block diagram schematically illustrates some exemplary embodiments according to this disclosure. Figure 1 The basic principles of information recommendation in the business system shown;

[0032] Figure 3 An information recommendation method according to some exemplary embodiments of the present disclosure is schematically illustrated in the form of a flowchart;

[0033] Figure 4 An information recommendation method according to some other exemplary embodiments of the present disclosure is schematically illustrated in the form of a flowchart;

[0034] Figure 5 In the form of a flowchart Figure 4 The information recommendation method shown is described in more detail;

[0035] Figure 6 In the form of a flowchart Figure 4 The information recommendation method shown is described in more detail;

[0036] Figure 7 The structure of an information recommendation device according to some exemplary embodiments of the present disclosure is schematically shown;

[0037] Figure 8 The structure of an information recommendation device according to some other exemplary embodiments of the present disclosure is schematically shown;

[0038] Figure 9 right Figure 8 The modules shown are described in more detail;

[0039] Figure 10 right Figure 8 The modules shown are described in more detail; and

[0040] Figure 11The structure of a computing device according to some exemplary embodiments of the present disclosure is illustrated schematically. The computing device includes the information recommendation apparatus described in the present disclosure, thereby enabling the implementation of the information recommendation method described in the present disclosure.

[0041] It should be noted that the contents shown in the accompanying drawings are merely illustrative and therefore need not be drawn to scale. Furthermore, identical or similar features are indicated by the same reference numerals throughout all the drawings. Detailed Implementation

[0042] The following description provides specific details of various exemplary embodiments of this disclosure so that those skilled in the art can fully understand and implement the various exemplary embodiments of this disclosure.

[0043] First, some terms used in the exemplary embodiments of this disclosure will be explained to facilitate understanding by those skilled in the art:

[0044] Same information: Information involving the same recommended object. Generally, a large number of pieces of information may exist in an information recommendation database, which, although having different information identifiers (i.e., information IDs), involve the same recommended object in terms of content. In this disclosure, such information is considered the same information. In practice, users are also more likely to consider information involving the same recommended object as the same information, rather than information with the same information ID. Therefore, it should be understood that in this disclosure, information involving the same recommended object is considered the same information, but they do not necessarily have the same information ID.

[0045] Click-through rate (CTR): The ratio of the number of times a piece of information is clicked to the number of times it is displayed within a statistical period. CTR is a percentage that can be used to measure the level of attention a piece of information receives. The CTR of an information can be calculated as: CTR = (Number of clicks / Number of impressions) × 100%. For example, if an information is displayed 10 times and clicked 5 times within a statistical period, then the CTR of that information is 50%.

[0046] Conversion Rate (CR): Within a statistical period, the ratio of the number of times a conversion is completed based on a piece of information to the number of times that information is clicked. CR is also a percentage, used to measure the revenue-generating ability of an information. A conversion is an action that an information provider expects from a user based on a piece of information. Depending on the situation, conversion can take different forms. For example, a conversion could be completing a purchase of a product, or browsing a specific page and completing a registration, etc. The CR of an information can be calculated as: CR = (Number of conversions completed / Number of clicks) × 100%. For example, if 10 users click on an information within a statistical period, and 5 of them complete a conversion (i.e., complete the action expected by the information provider based on that information), then the CR of that information is 50%.

[0047] Predicted Click-Through Rate (pCTR): In an information recommendation system, the pCTR prediction model calculates the probability that a user will click on each piece of information in the candidate information set. Therefore, pCTR can be used to predict the probability that a piece of information will be clicked by a user. As a non-limiting example, the pCTR prediction model can calculate the probability that a piece of information will be clicked by a user based on information features (such as historical click-through rates, text, format, images, etc.) and user features (such as gender, age, interests, browsing history, etc.).

[0048] Predicted Conversion Rate (pCVR): In information recommendation systems, the conversion rate prediction model calculates the probability that a user will convert after clicking on each piece of information in the candidate information set. Therefore, pCVR can be used to predict the probability that a user who clicks on an item will subsequently complete a conversion. As a non-limiting example, the conversion rate prediction model can calculate the probability that a user will complete a conversion after clicking on an item based on information characteristics (such as historical conversion rates, text, format, images, etc.) and user characteristics (such as gender, age, interests, etc.).

[0049] Traffic Value (GMV): In an information recommendation system, it is defined as the sum of the bids from information providers for each conversion within a statistical period. The traffic value of a piece of information can be calculated as: GMV = Number of Conversions × Information Bid. Therefore, the traffic value of a piece of information reflects the revenue that information brings to the information recommendation system, taking into account both traffic and information bid. Thus, the traffic value of information is related to the revenue of the information recommendation system.

[0050] Expected revenue per thousand impressions (eCPM): This is calculated for each piece of information in the candidate information set, based on the information provider's bid and the corresponding pCTR and pCVR. In other words, the expected revenue per thousand impressions for a piece of information can be calculated as: eCPM = pCTR × pCVR × Information Bid. Therefore, the expected revenue per thousand impressions reflects the anticipated revenue that the information can bring to the information recommendation system. Thus, the expected revenue per thousand impressions is directly related to the expected revenue of the information recommendation system.

[0051] Determinant Point Process (DPP): A process that transforms complex probability model calculations into simple matrix determinant calculations.

[0052] Information freshness: The similarity of a piece of information to a set of previously displayed information. Information freshness reflects the effectiveness of the displayed information together with the information recently displayed to the user; a higher value indicates a better display set.

[0053] Currently, in related technologies, to control the repeated recommendation of the same information to users, business applications generally adopt a freshness strategy. Specifically, the business can strictly limit the number of times the same information is repeatedly recommended to users within a predetermined time period based on the user's information display history. For example, the same information can be limited to being recommended / displayed only once every 10 minutes or 1 hour, thus giving different information a chance to be recommended and displayed. Although the freshness strategy can greatly reduce the repeated recommendation / display of the same information, thereby improving the user experience, it still suffers from the problems mentioned earlier.

[0054] See Figure 1 This illustration demonstrates a general application scenario of a business system according to some exemplary embodiments of this disclosure. For example... Figure 1 As shown, the business system 100 may include a business terminal 110 and one or more terminal devices 130, wherein the business terminal 110 is capable of communicating with one or more terminal devices 130 via a network 120. Thus, the business terminal 110 can provide services to users of one or more terminal devices 130, including but not limited to information recommendation.

[0055] The business terminal 110 stores and runs computer instructions capable of executing the various methods described in this disclosure. It can be a single server, a server cluster, or a cloud server. The server can be a standalone physical server, a server cluster consisting of multiple physical servers, or a distributed system. It can also be a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. The business terminal 110 and the terminal device 130 can communicate via wired or wireless communication, and this disclosure does not impose any limitations on this.

[0056] Terminal device 130 can be any type of suitable mobile device, including but not limited to: mobile computers (e.g., but not limited to, Microsoft® Surface® devices, personal digital assistants (PDAs), laptops, notebook computers, tablet computers such as Apple iPad™, netbooks, etc.), smartphones (e.g., but not limited to, Microsoft Windows® phones, Apple iPhones, phones implementing the Google® Android™ operating system, Palm® phones, Blackberry® phones, etc.), wearable computing devices (e.g., but not limited to, smartwatches and wristbands, head-mounted devices, including smart glasses such as Google® Glass™, etc.), or other types of mobile devices. Furthermore, in some exemplary embodiments, terminal device 130 can also be any type of suitable fixed device, including but not limited to, desktop computers, game consoles, smart TVs, smart speakers, projectors, set-top boxes, outdoor advertising displays, in-vehicle advertising displays, etc.

[0057] Network 120 may include, but is not limited to, any combination of a local area network (LAN), a wide area network (WAN), a personal area network (PAN), and / or a communication network such as the Internet. Each of the service terminal 110 and one or more terminal devices 130 may include at least one communication interface (not shown) capable of communicating via network 120. Such a communication interface may be one or more of the following: any type of network interface (e.g., a network interface card (NIC)), wired or wireless (such as IEEE 802.11 Wireless LAN (WLAN)) wireless interface, Wi-MAX interface, Ethernet interface, Universal Serial Bus (USB) interface, cellular network interface, Bluetooth™ interface, Near Field Communication (NFC) interface, etc.

[0058] exist Figure 1In the business system 100 shown, users can perform certain application operations through terminal device 130, such as, but not limited to, browsing information, reading novels, playing videos, or playing games. In these scenarios, users can send requests for information recommendations to the business terminal 110 via network 120 through terminal device 130. After receiving the request from terminal device 130, business terminal 110 prepares the information to be recommended and sends it to terminal device 130 via network 120. Then, terminal device 130 can display the received information to the user. In other examples, the requests for information recommendations in the above scenarios can also be automatically sent by a third-party application in terminal device 130, for example, requests for information recommendations can be automatically sent at predetermined time intervals.

[0059] See Figure 2 It schematically illustrates, in block diagram form, some exemplary embodiments according to this disclosure. Figure 1 The basic principle of information recommendation in the business system shown.

[0060] like Figure 2 As shown and in conjunction with reference Figure 1 Terminal device 130 can send a request for information recommendation to service terminal 110 via network 120. Then, terminal device 130 can wait for feedback from service terminal 110, and when it receives candidate information sent from service terminal 110, it can display the candidate information to the user of terminal device 130.

[0061] The service terminal 110 receives an information recommendation request from the terminal device 130, and upon receiving the request, generates a candidate information queue for display on the terminal device. The candidate information queue may include multiple candidate information items, and these items can be sorted according to a predetermined metric. Generally, the revenue metric corresponding to the candidate information can be used as the predetermined metric. As previously explained, the eCPM and GMV of an information item are associated with the expected revenue and revenue of the information recommendation system, respectively. Therefore, in some exemplary embodiments, eCPM or GMV can be used as the predetermined metric to sort the candidate information from high to low according to the predetermined metric. However, it should be understood that it is also possible to sort the candidate information in the candidate information queue based on any other suitable metric, and this disclosure does not limit this.

[0062] For the generated candidate information queue, the service terminal 110 will determine whether the first candidate information in the first ranking position is repeatedly recommended. As a non-limiting example, if the time interval between the last time the first candidate information was displayed on the same terminal device 130 and the current time is less than a predetermined time period, it can be determined that the first candidate information is repeatedly recommended. For example, the predetermined time period can be 10 minutes or 1 hour. If it is determined that the first candidate information is not repeatedly recommended, then the service terminal 110 will send the first candidate information to the terminal device 130 via the network 120. If it is determined that the first candidate information is repeatedly recommended, then the service terminal 110 will further determine whether it is necessary to intervene in such ranking. In some exemplary embodiments, the service terminal 110 can determine the estimated attractiveness of the first candidate information to the user, and use this as a basis to determine whether it is necessary to intervene in such ranking. Estimated attractiveness refers to the estimated probability that the user is interested in the information and may click or convert. As a non-limiting example, the pCTR of the first candidate information can be used as its estimated attractiveness. For example, when the pCTR of the first candidate information is higher than or equal to 50%, it indicates that it still has a relatively high probability of attracting user clicks. Therefore, the business terminal 110 will not intervene in such ranking, but will send the first candidate information to the terminal device 130 via the network 120. When the pCTR of the first candidate information is less than 50%, it indicates that it no longer has sufficient attractiveness. Therefore, the business terminal 110 will re-rank the candidate information in the candidate information queue. In addition, it should be understood that it is possible to use any other suitable indicator as the estimated attractiveness of the first candidate information, and this disclosure is not limited in this regard. As a non-limiting example, the pCVR of the first candidate information can also be used as its estimated attractiveness.

[0063] When it is necessary to reorder the candidate information in the candidate information queue, the service terminal 110 will perform probabilistic modeling based on the information freshness of each candidate information, that is, calculate the information freshness of each candidate information. Then, considering both the eCPM and information freshness of each candidate information, for example, multiplying the two, the candidate information in the candidate information queue is reordered. In some other exemplary embodiments, the GMV and information freshness of each candidate information can also be considered comprehensively, for example, multiplying the two, in order to reorder the candidate information in the candidate information queue. Then, the service terminal 110 can send the reordered candidate information in the first sorted position to the terminal device 130 via the network 120, so that the terminal device 130 can display it to the user.

[0064] In the information recommendation process based on the above principles, on the one hand, the conditions for intervening in information that is repeatedly recommended are multiple. It is necessary not only to determine that candidate information is repeatedly recommended, but also to dynamically determine whether it should continue to be displayed based on the estimated attractiveness of the candidate information (such as pCTR or pCVR). On the other hand, the intervention process takes into account both the freshness of information and the expected benefits, thereby balancing user experience and business system benefits.

[0065] See now Figure 3 The flowchart schematically illustrates an information recommendation method 300 according to some exemplary embodiments of the present disclosure. As a non-limiting example, the information recommendation method 300 can be applied to... Figure 1 , Figure 2 The business terminal 110 of the business system 100 shown provides information recommendation services. For example... Figure 3 As shown, the information recommendation method 300 may include steps 310, 320, and 330:

[0066] Step 310: In response to the information recommendation request received from the terminal device, a candidate information queue is generated, wherein the candidate information queue includes first candidate information at a first sorting position;

[0067] Step 320: When the first candidate information involves the same recommended object as the recommendation information recommended to the terminal device within a predetermined time period prior to the current time, determine the estimated attractiveness of the first candidate information, where the estimated attractiveness is the probability that the user of the terminal device will perform a desired action based on the first candidate information; and

[0068] Step 330: When the estimated attractiveness is greater than or equal to a predetermined attractiveness threshold, the first candidate information is recommended.

[0069] In step 310, the generated candidate information queue may include multiple candidate information items, and these candidate information items can be sorted according to a predetermined indicator. As mentioned earlier, the revenue indicator corresponding to the candidate information can be selected as the predetermined indicator. For a piece of information, its eCPM and GMV are both related to the revenue that the information can bring to the information recommendation system. The higher the eCPM, the higher the expected revenue of the information recommendation system; the higher the GMV, the higher the revenue of the information recommendation system. Therefore, for the candidate information queue, as a non-limiting example, eCPM or GMV can be used as the predetermined indicator to sort the candidate information from high to low. For example, in the non-limiting example of using eCPM as the predetermined indicator, when sorting each candidate information item, the eCPM value of each candidate information item can be calculated separately. For example, the eCPM value of each candidate information item can be calculated as: eCPM = pCTR × pCVR × information bid. Then, based on the calculation results of the eCPM values ​​of each candidate information item, they are sorted in descending order, so that the first candidate information item located in the first sorting position has the largest eCPM value. However, it should be understood that it is possible to rank the candidate information in the candidate information queue based on any other suitable metric, and this disclosure does not limit this.

[0070] In step 320, as a non-limiting example, the existence of duplicate recommendations for the first candidate information can be determined based on the time interval between the display of the same information to the same terminal device 130. As previously explained, in this disclosure, "same information" refers to information involving the same recommended object. Therefore, although some information may have different information identifiers (i.e., information IDs), the recommended object they involve is the same in terms of content. In this disclosure, such information is also considered the same information. Therefore, when the first candidate information and the recommended information recommended to the same terminal device 130 before the current moment involve the same recommended object, they are considered the same information, and if the time interval between the current moment and the moment when the previously recommended information was recommended is within a predetermined time period, it can be determined that the first candidate information is a duplicate recommendation. As a non-limiting example, the predetermined time period may be 10 minutes or 1 hour. Furthermore, the estimated attractiveness refers to the estimated probability that a user is interested in the information and may perform a desired action. In this disclosure, as a non-limiting example, the probability that a user performs a desired action may refer to the probability that a user, based on the information, will subsequently click or subsequently perform a conversion behavior. Therefore, the pCTR of the first candidate information can be used as its estimated attractiveness. For example, when the pCTR of the first candidate information is higher than or equal to 50%, it indicates that it still has a relatively high probability of attracting user clicks; while when the pCTR of the first candidate information is less than 50%, it indicates that it no longer has sufficient attractiveness. Furthermore, it should be understood that it is also possible to use any other suitable metric as the estimated attractiveness of the first candidate information, and this disclosure does not limit this. For example, the pCVR of the first candidate information can also be used as its estimated attractiveness.

[0071] In step 330, when the estimated attractiveness of the first candidate information is greater than or equal to a predetermined attractiveness threshold, the information recommendation method 300 will recommend the first candidate information. As described above, when using pCTR or pCVR as the estimated attractiveness, the predetermined attractiveness threshold can be 50%.

[0072] According to the information recommendation method 300 disclosed herein, when candidate information is repeatedly recommended, the method further dynamically determines whether the repeated recommendation should continue based on its estimated attractiveness. This allows for a moderate amount of repeated recommendation of the same information, thereby reducing the overall repetition rate of the same information and improving the user experience, while also increasing the display of high-quality information and thus increasing the revenue of the information recommendation platform.

[0073] See Figure 4 The flowchart schematically illustrates an information recommendation method 400 according to some other exemplary embodiments of the present disclosure. Figure 4As shown, the information recommendation method 400 may include steps 410, 420, 430, 440, 450, 460, and 470:

[0074] Steps 410 and 410 are respectively with Figure 3 Steps 310 and 320 in the information recommendation method 300 shown are the same and will not be repeated here;

[0075] Step 430: Determine whether the estimated attractiveness of the first candidate information is less than a predetermined attractiveness threshold; if the estimated attractiveness of the first candidate information is not less than the predetermined attractiveness threshold, then the information recommendation method 400 will execute step 470 to recommend the first candidate information in the candidate information queue; if the estimated attractiveness of the first candidate information is less than the predetermined attractiveness threshold, then the information recommendation method 400 will execute steps 440, 450, and 460.

[0076] Step 440: Determine the information freshness of each candidate information in the candidate information queue. The information freshness is the overall similarity between each candidate information and at least one piece of information that has been displayed by the terminal device before the current time.

[0077] Step 450: Based on the information freshness of each candidate information in the candidate information queue and the corresponding expected revenue per thousand exposures, reorder each candidate information to obtain the first reordered candidate information of the candidate information queue.

[0078] Step 460: Recommend the first rearranged candidate information of the candidate information queue.

[0079] As explained above, information freshness represents the overall similarity of a piece of information relative to a set that includes at least one piece of previously displayed information. Therefore, information freshness reflects the effectiveness of the information presented, along with the set of information most recently shown to the user. In this disclosure, as a non-limiting example, information freshness can be obtained based on a determinant point process (DPP). Specifically, based on several recently displayed pieces of information, the similarity of each candidate piece of information in the candidate information queue is calculated relative to these several pieces of information. Then, a similarity matrix is ​​constructed based on the obtained similarity, and the determinant of this similarity matrix is ​​calculated using the DPP process, thereby obtaining the information freshness of each candidate piece of information. It should be understood that a higher information freshness value means a better freshness effect for the corresponding candidate piece of information presented, along with the set of information most recently shown. Therefore, sorting the candidate pieces of information considering their information freshness allows for the selection of high-quality information from the candidate information.

[0080] See Figure 5 It uses a flowchart to illustrate... Figure 4The determination of information freshness is described in more detail in step 440 of the information recommendation method 400 shown. For example... Figure 5 As shown, Figure 4 Step 440 in the information recommendation method 400 shown may include steps 4401, 4402, 4403, and 4404:

[0081] Step 4401: Obtain the previous status of the corresponding terminal device from the current time. n The information already displayed in this session includes... n It is an integer greater than 0;

[0082] Step 4402: Based on the displayed information, construct... n × n The similarity matrix;

[0083] Step 4403: Traverse each candidate message in the candidate message queue and construct an augmented similarity matrix corresponding to each candidate message; and

[0084] Step 4404: Calculate the determinant of the augmented similarity matrix of each candidate information in the candidate information queue, and use the values ​​obtained from the determinant calculation as the information freshness of each candidate information.

[0085] In the recommendation method based on the information disclosed herein n In practice, any integer greater than 0 can be used as needed, as long as its value does not affect the calculation speed and thus reduce the user experience. n × n Similarity matrix S In the matrix, each element s ij The values ​​are as follows:

[0086] Formula 1

[0087] in, α ij Indicates the first i The information already displayed is related to the first item. j The similarity of the displayed information is 0 < α ij ≤1, and i and j All are integers, 0 < i ≤ n ,0< j ≤ n As a non-restrictive example, the first... i The information already displayed is related to the first item. j Similarity of the displayed information α ij It can be calculated as ,in, m ij An integer greater than or equal to 0 indicates the first... i The information already displayed is related to the first item. j The direct, non-repeating distance between the displayed information items (also known as Manhattan distance).

[0088] As a non-restrictive example, direct non-repeating distance m ij The calculation method is as follows: First, let its initial value be 0; then, determine the first... i The information already displayed is related to the first item. j The number of items showing differences between the displayed information can be used as the direct non-repeating distance. m ij The value of . For example, if the . i The information already displayed is related to the first item. j The information displayed comes from different information providers, so the distance should be used to avoid repetition. m ij The value can be incremented by 1; furthermore, if the first... i The information already displayed is related to the first item. j The information already displayed involves different industries, so we will directly avoid repeating any information. m ij The value can be increased by 1. Therefore, if the first... i The information already displayed is related to the first item. j The greater the differences between the displayed information items, the greater the direct non-repetition distance between them. m ij The larger the similarity, the greater the similarity between them. α ij The closer the similarity is to 0, the closer it is to 0; conversely, the closer the similarity is to 0, the closer it is to 0. α ij The closer it is to 1. It should be understood that the above text regarding the first... i The information already displayed is related to the first item. j The direct, non-repeating distance of the displayed information m ij It refers to the similarity matrix. S The description is as follows, and it is easy to recognize that direct non-repeating distance can be used to calculate the similarity between any two pieces of information.

[0089] Furthermore, as mentioned above, in the similarity matrix S In the middle, s ij =s ji Therefore, the similarity matrix S The matrix form is as follows:

[0090] Formula 2

[0091] In step 4403, when there are candidate information queues M When selecting candidate information, the process involves selecting one candidate information from the candidate information queue. v x Calculate its relationship with the previous one respectively. n Similarity of the displayed information v x1 , v x2 , ... , v xn ,in x are integers and 1 ≤ x ≤ M As a non-limiting example, the similarity between two pieces of information can be calculated based on the direct, non-repeating distance between them. Then, based on the calculated similarity... v x1 , v x2 , ... , v xn This can be used to construct the candidate information. v x Compared to n Similarity vectors of the displayed information V x = [ v x1 , v x2 , ... , v xn Therefore, for this candidate information in the candidate information queue... v x It is possible to construct its corresponding ( n +1)×( n Augmented similarity matrix (+1) Its matrix form is as follows:

[0092] Formula 3

[0093] in x are integers and 1 ≤ x ≤ M Repeat the above process to iterate through all candidate information in the queue. M Each candidate piece of information is constructed with... M One-to-one correspondence of candidate information M An augmented similarity matrix.

[0094] In step 4404, it can be based on and M One-to-one correspondence of candidate information M For each augmented similarity matrix, calculate the value of its determinant.D x The value of the determinant D x Can be used as candidate information v x The freshness of information.

[0095] In some cases, n The displayed information in a list of already displayed information may be completely identical to the candidate information in the candidate information queue, for example, when no freshness strategy is adopted to control the repeated display of the same information. In such cases, the similarity matrix... S and augmented similarity matrix There will be at least two columns with exactly the same value, which will cause the value of the determinant to be... D x A similarity score of 0 ultimately leads to an inability to obtain accurate information freshness about the candidate information. To avoid this problem, one approach is to construct a similarity matrix... S and augmented similarity matrix During the process, except for matrix elements s 11 , s 22 , ..., s nn , s (n+1)(n+1) Apart from the matrix elements, all other matrix elements have values ​​less than 1. As a non-restrictive example, this is used when constructing a similarity matrix. S Then, Equation 1 above can be modified as follows:

[0096] Equation 4

[0097] in, β It is a real number less than 1 but very close to 1. As an example, β It can be 0.9995. According to Equation 4, when the calculated similarity is... α ij Value greater than β When, matrix elements s ij The value is set to β The value of .

[0098] See Figure 6 It uses a flowchart to illustrate... Figure 4 Step 450 of the information recommendation method 400 shown is described in more detail. For example... Figure 6 As shown, Figure 4 Step 450 in the information recommendation method 400 shown may include steps 4501 and 4502:

[0099] Step 4501: Multiply the information freshness of each candidate information in the candidate information queue by the corresponding expected revenue per thousand impressions to obtain the corresponding recommendation index; and

[0100] Step 4502: Select the candidate information with the highest recommendation index from among the candidate information in the candidate information queue as the first rearranged candidate information.

[0101] Figure 6 The method described generates recommendation metrics for ranking and recommending candidate information based on a comprehensive consideration of information freshness and expected revenue per thousand impressions or traffic value, which reflects the desired revenue. This approach balances information freshness and platform revenue, minimizing the impact of the freshness strategy on platform revenue. However, it should be understood that other metrics reflecting platform revenue can also be used in conjunction with information freshness to generate corresponding recommendation metrics.

[0102] The information recommendation method disclosed herein has at least the following beneficial technical effects: First, by setting an estimated attractiveness index for information, it dynamically determines whether the same information should continue to be displayed, thereby allowing competitive, identical information to be displayed multiple times within a short period of time, and enabling the replacement of uncompetitive information to give display opportunities to other information; Second, the process of adjusting candidate information takes into account both consumption loss and information freshness index, thereby minimizing the adverse impact of freshness strategies on business revenue. In summary, the information recommendation method and apparatus disclosed herein, by dynamically allowing appropriate repeated display of information, can establish a mutually satisfactory balance between user experience and business revenue.

[0103] See Figure 7 This schematically illustrates the structure of an information recommendation device 700 according to some exemplary embodiments of the present disclosure. The information recommendation device 700 can be applied to the service terminal 110 and can be applied to various scenarios described in this disclosure. For example... Figure 7 As shown, the information recommendation device 700 may include: a candidate information queue generation module 710, an estimated attractiveness determination module 720, and a recommendation module 730.

[0104] The candidate information queue generation module 710 is configured to generate a candidate information queue in response to an information recommendation request received from a terminal device, wherein the candidate information queue includes first candidate information at a first sorting position. The estimated attractiveness determination module 720 is configured to determine the estimated attractiveness of the first candidate information when it relates to the same recommended object as recommendation information recommended to the terminal device within a predetermined time period prior to the current time; the estimated attractiveness is the probability that the user of the terminal device will perform a desired action based on the first candidate information. The recommendation module 730 is configured to recommend the first candidate information in the candidate information queue when the estimated attractiveness is greater than or equal to a predetermined attractiveness threshold.

[0105] See Figure 8 This schematically illustrates the structure of an information recommendation device 800 according to some other exemplary embodiments of the present disclosure. Similarly, the information recommendation device 800 can be applied to the service terminal 110 and can be applied to various scenarios described in this disclosure. For example... Figure 8 As shown, the information recommendation device 800 may include: a candidate information queue generation module 810, an estimated attractiveness determination module 820, an estimated attractiveness judgment module 830, an information freshness determination module 840, a re-sorting module 850, and a recommendation module 860.

[0106] The candidate information queue generation module 810 and the estimated attractiveness determination module 820 are respectively connected with... Figure 7 The candidate information queue generation module 710 and the estimated attractiveness determination module 720 in the information recommendation device 700 shown are the same and will not be described again here. The estimated attractiveness judgment module 830 is configured to: determine whether the estimated attractiveness determined by the estimated attractiveness determination module 820 is less than a predetermined attractiveness threshold. The information freshness determination module 840 is configured to: determine the information freshness of each candidate information in the candidate information queue, wherein the information freshness is the overall similarity between each candidate information and at least one piece of information displayed by the terminal device before the current time. The reordering module 850 is configured to: reorder each candidate information based on the information freshness of each candidate information in the candidate information queue and the corresponding expected revenue per thousand exposures, to obtain the first candidate information of the reordered candidate information queue. The recommendation module 860 is configured to: recommend the first candidate information of the candidate information queue when the estimated attractiveness is not less than the predetermined attractiveness threshold, and recommend the first reordered candidate information of the candidate information queue when the estimated attractiveness is less than the predetermined attractiveness threshold.

[0107] See Figure 9 It schematically shows Figure 8 The structure of the freshness determination module 840. For example... Figure 9As shown, the freshness determination module 840 may include: a displayed information acquisition submodule 8401, a similarity matrix generation submodule 8402, an augmented similarity matrix generation submodule 8403, and an information freshness generation submodule 8404.

[0108] The displayed information acquisition submodule 8401 is configured to: acquire information about the corresponding terminal device from the current moment. n The information already displayed in this session includes... n The integer is greater than 0. The similarity matrix generation submodule 8402 is configured to: construct based on the displayed information. n × n Similarity matrix S The similarity matrix S Each matrix element ,in α ij Indicates the first i The information already displayed is related to the first item. j The similarity of the displayed information is 0 < α ij ≤1, and i and j All are integers, 0 < i ≤ n ,0< j ≤ n , and s ij =s ji The augmented similarity matrix generation submodule 8403 is configured to: traverse each candidate piece of information in the candidate information queue and construct an augmented similarity matrix corresponding to each candidate piece of information. ,in, S This is the similarity matrix. V It is the similarity vector of the candidate information relative to the already displayed information. V =[ v 1 , v 2 , ... , v n ],in v 1 , v 2 , ... , v n This candidate information is respectively with n The similarity of the displayed information, and v 1 , v 2 , ... , v nIn the matrix, each element is greater than 0 and less than or equal to 1, and T represents the transpose of the matrix. The information freshness generation submodule 8404 is configured to: perform determinant calculation on the augmented similarity matrix of each candidate information in the candidate information queue, and use the values ​​obtained from the determinant calculation as the information freshness of each candidate information.

[0109] See Figure 10 It schematically shows Figure 8 The structure of the reordering module 850 in [the text]. For example... Figure 10 As shown, the reordering module 850 may include a recommendation metric generation submodule 8501 and a first reordering candidate information determination submodule 8502. The recommendation metric generation submodule 8501 is configured to multiply the information freshness of each candidate information in the candidate information queue by its corresponding expected revenue per thousand exposures to obtain the corresponding recommendation metric. The first reordering candidate information determination submodule 8502 is configured to use the candidate information with the highest recommendation metric among all candidate information in the candidate information queue as the first reordering candidate information.

[0110] It should be understood that the above regarding Figures 7 to 10 The various modules and submodules described can be implemented in hardware or in hardware in combination with software and / or firmware. For example, these modules and submodules can be implemented as computer-executable code / instructions configured to execute in one or more processors and stored in a computer-readable storage medium. Alternatively, these modules and submodules can be implemented as hardware logic / circuit. For example, in some embodiments, one or more of these modules can be implemented together in a system-on-a-chip (SoC). An SoC may include an integrated circuit chip (which includes a processor (e.g., a central processing unit (CPU), microcontroller, microprocessor, digital signal processor (DSP), etc.), memory, one or more communication interfaces, and / or one or more components in other circuitry), and may optionally execute received program code and / or include embedded firmware to perform functions. It should also be understood that the various modules and submodules described above relate to those described above regarding... Figure 3 The description of information recommendation method 300, and about Figure 4 , Figure 5 , Figure 6 The operations of each relevant step in the information recommendation method 400 are described, and therefore will not be repeated here.

[0111] Figure 11 A schematic block diagram of an exemplary computing device 1100 according to some embodiments of the present disclosure is shown. This exemplary computing device 1100 may represent... Figure 1 The business terminal 110 shown is an exemplary computing device 1100 that can be used in various scenarios described in this disclosure.

[0112] The computing device 1100 may include at least one processor 1102, memory 1104, multiple communication interfaces 1106, display device 1108, other input / output (I / O) devices 1110, and one or more mass storage devices 1112 that are capable of communicating with each other, such as via a system bus 1114 or other suitable means.

[0113] Processor 1102 may be a single processing unit or multiple processing units, and all processing units may include single or multiple computing units or multiple cores. Processor 1102 may be implemented as one or more microprocessors, microcomputers, microcontrollers, digital signal processors, central processing units, state machines, logic circuits, and / or any device that manipulates signals based on operating instructions. Among other capabilities, processor 1102 may be configured to acquire and execute computer-readable instructions stored in memory 1104, mass storage device 1112, or other computer-readable media, such as program code of operating system 1116, program code of application program 1118, program code of other program 1120, etc.

[0114] Memory 1104 and mass storage device 1112 are examples of computer storage media for storing instructions executed by processor 1102 to perform the various functions described above. For example, memory 1104 may generally include both volatile and non-volatile memory (e.g., RAM, ROM, etc.). Furthermore, mass storage device 1112 may generally include hard disk drives, solid-state drives, removable media, including external and removable drives, memory cards, flash memory, floppy disks, optical disks (e.g., CDs, DVDs), storage arrays, network-attached storage, storage area networks, etc. Memory 1104 and mass storage device 1112 may be collectively referred to as memory or computer storage media in this disclosure, and may be non-transitory media capable of storing computer-readable, processor-executable program instructions as computer-executable code, which may be executed by processor 1102 as a specific machine configured to perform the operations and functions described in the examples of this disclosure.

[0115] Multiple program modules can be stored on a mass storage device 1112. These program modules include an operating system 1116, one or more application programs 1118, other programs 1120, and program data 1122, and they can be executed by a processor 1102. Examples of such application programs or program modules may include, for example, computer program logic (e.g., computer-executable code or instructions) for implementing the following components / functions: candidate information queue generation modules 710, 810, estimated attractiveness determination modules 720, 820, and recommendation modules 730, 860; estimated attractiveness judgment module 830, information freshness determination module 840, and reordering module 850; displayed information acquisition submodule 8401, similarity matrix generation submodule 8402, augmented similarity matrix generation submodule 8403, and information freshness generation submodule 8404; and recommendation index generation submodule 8501 and first reordered candidate information determination submodule 8502.

[0116] Although Figure 11 The modules are illustrated as being stored in memory 1104 of computing device 1100, but modules 1116, 1118, 1120, and 1122, or portions thereof, may be implemented using any form of computer-readable medium accessible by computing device 1100. As used in this disclosure, "computer-readable medium" includes at least two types of computer-readable media: computer-readable storage media and communication media.

[0117] Computer-readable storage media include non-transitory, volatile and non-volatile, removable and non-removable media implemented by any method or technology for storing information, such as computer-readable instructions, data structures, program modules, or other data. Computer-readable storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, Digital Universal Disc (DVD), or other optical storage devices, magnetic cassettes, magnetic tapes, disk storage devices or other magnetic storage devices, or any other non-transfer medium that can be used to store information for access by a computing device.

[0118] In contrast, communication media can embody computer-readable instructions, data structures, program modules, or other data in modulated data signals such as carrier waves or other transmission mechanisms. Computer-readable storage media as defined in this disclosure do not include communication media.

[0119] The computing device 1100 may also include one or more communication interfaces 1106 for exchanging data with other devices, such as via a network, direct connection, etc. The communication interface 1106 can facilitate communication across various network and protocol types, including wired networks (e.g., LAN, cable, etc.) and wireless networks (e.g., WLAN, cellular, satellite, etc.), the Internet, etc. The communication interface 1106 can also provide communication with external storage devices (not shown), such as storage arrays, network-attached storage, storage area networks, etc.

[0120] In some examples, computing device 1100 may include display device 1108, such as a monitor, for displaying information and images. Other I / O devices 1110 may be devices that receive various inputs from a user and provide various outputs to the user, including but not limited to touch input devices, gesture input devices, cameras, keyboards, remote controls, mice, printers, audio input / output devices, and so on.

[0121] The terminology used in this disclosure is for the purpose of describing embodiments thereof and is not intended to limit the disclosure. As used herein, the singular forms “a,” “an,” and “the” are intended to include the plural forms as well, unless the context clearly indicates otherwise. It should also be understood that the terms “comprising” and “including,” as used herein, refer to the presence of the stated feature but do not exclude the presence of one or more other features or the addition of one or more other features. As used herein, the term “and / or” includes any and all combinations of one or more of the associated listed items. It will be understood that although the terms “first,” “second,” “third,” etc., may be used in this disclosure to describe various features, these features should not be limited by these terms. These terms are used only to distinguish one feature from another.

[0122] Unless otherwise defined, all terms used in this disclosure (including technical and scientific terms) shall have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure pertains. It shall also be understood that terms such as those defined in commonly used dictionaries shall be interpreted as having the meaning consistent with their meaning in the relevant field and / or the context of this specification, and shall not be interpreted in an idealized or overly formal sense unless expressly defined in this disclosure.

[0123] In the description of this specification, the terms "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to a specific feature, structure, material, or characteristic described in connection with that embodiment or example that is included in at least one embodiment or example of this disclosure. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0124] This disclosure describes various techniques in the general context of software, hardware, components, or program modules. Generally, these modules include routines, programs, objects, elements, components, data structures, etc., that perform a specific task or implement a specific abstract data type. As used in this disclosure, the terms "module," "function," and "component" generally refer to software, firmware, hardware, or a combination thereof. The techniques described in this disclosure are platform-independent, meaning that these techniques can be implemented on a variety of computing platforms with various processors.

[0125] It should be understood that the order of the steps in the methods described in this disclosure is merely exemplary and not restrictive. Therefore, the steps in the methods described in this disclosure do not necessarily have to be performed in the described order, but may be performed in a different order as needed, or may include any suitable additional steps.

[0126] It should be understood that the logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable storage medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples of computer-readable storage media may include, for example, electrical connections having one or more wires (electronic devices), portable computer disk drives (magnetic devices), random access memory, read-only memory, erasable programmable read-only memory or flash memory, fiber optic devices, and compact disc read-only memory. Alternatively, the computer-readable storage medium could be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0127] It should be understood that various parts of this disclosure can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (FPGAs), field-programmable gate arrays (FPGAs), etc.

[0128] Those skilled in the art will understand that all or part of the steps of the methods described in the above embodiments can be implemented by hardware associated with program instructions. The program can be stored in a computer-readable storage medium, and when executed, the program includes performing one or a combination of the steps of the method embodiments.

[0129] Although this disclosure has been described in detail with reference to some embodiments, it is not intended to be limited to the specific forms set forth herein. Rather, the scope of this disclosure is defined only by the appended claims.

Claims

1. An information recommendation method, comprising: In response to an information recommendation request received from a terminal device, a candidate information queue is generated, the candidate information queue including first candidate information at a first sorting position; When the first candidate information involves the same recommended object as the recommendation information recommended to the terminal device within a predetermined time period before the current time, the estimated attractiveness of the first candidate information is determined. The estimated attractiveness is the probability that the user of the terminal device will take the desired action based on the first candidate information. The estimated attractiveness includes the estimated click-through rate or the estimated conversion rate of the first candidate information. When the estimated attractiveness is greater than or equal to the predetermined attractiveness threshold, the first candidate information is recommended; When the estimated attractiveness is less than the predetermined attractiveness threshold: Obtain the previous state of the terminal device from the current moment. n The information already displayed in this session includes... n It is an integer greater than 0; Based on the information already presented, construct n × n Similarity matrix S Among them, the similarity matrix S Each matrix element ,in α ij Indicates the first i The information already displayed is related to the first item. j The similarity of the displayed information is 0 < α ij ≤1, i and j All are integers, 0 < i ≤ n ,0< j ≤ n , and s ij =s ji ; Traverse each candidate information in the candidate information queue and construct an augmented similarity matrix corresponding to each candidate information. ,in, S It is the similarity matrix, V It is the similarity vector of the candidate information relative to the already displayed information. V =[ v 1 , v 2 , ... , v n ],in v 1 , v 2 , ... , v n The aforementioned candidate information is respectively with n The similarity of the information already presented in the above clause, and v 1 , v 2 , ... , v n For each element in the matrix that is greater than 0 and less than or equal to 1, T represents the transpose of the matrix. The augmented similarity matrix of each candidate information in the candidate information queue is calculated by performing a determinant calculation, and each value obtained by the determinant calculation is used as the information freshness of each candidate information. Based on the information freshness and corresponding expected revenue per thousand impressions of each candidate information in the candidate information queue, the candidate information is reordered to obtain the first reordered candidate information of the candidate information queue; and Recommend the first rearrangement candidate information.

2. The information recommendation method as described in claim 1, wherein, In response to an information recommendation request received from a terminal device, a candidate information queue is generated, the candidate information queue including first candidate information at a first sorting position, including: The candidate information is sorted based on the expected revenue per thousand exposures to generate the candidate information queue, wherein the first candidate information has the highest expected revenue per thousand exposures.

3. The information recommendation method as described in claim 1, wherein, The similarity between two pieces of information is calculated as equal to ,in, m An integer greater than or equal to 0 represents the direct, non-repeating distance between the two pieces of information.

4. The information recommendation method as described in claim 1, wherein, In the similarity matrix, when i ≠ j When, 0 < α ij <1.

5. The information recommendation method as described in claim 1, wherein, The step of reordering the candidate information based on its freshness and corresponding expected revenue per thousand impressions in the candidate information queue includes: The information freshness of each candidate information in the candidate information queue is multiplied by the corresponding expected revenue per thousand exposures to obtain the corresponding recommendation index. The candidate information with the highest recommendation index among all candidate information in the candidate information queue is selected as the first rearranged candidate information.

6. An information recommendation device, comprising: A candidate information queue generation module is configured to generate a candidate information queue in response to an information recommendation request received from a terminal device, wherein the candidate information queue includes first candidate information at a first sorting position; An estimated attractiveness determination module is configured to: when the first candidate information involves the same recommended object as the recommendation information recommended to the terminal device within a predetermined time period before the current time, determine the estimated attractiveness of the first candidate information, wherein the estimated attractiveness is the probability that the user of the terminal device will perform a desired action based on the first candidate information, and the estimated attractiveness includes the estimated click-through rate or the estimated conversion rate of the first candidate information; The recommendation module is configured to recommend the first candidate information when the estimated attractiveness is greater than or equal to a predetermined attractiveness threshold. The information freshness determination module is configured to: when the estimated attractiveness is less than the attractiveness threshold, determine the information freshness of each candidate information in the candidate information queue, wherein the information freshness is the overall similarity between each candidate information and at least one piece of information that has been displayed by the terminal device before the current time. The reordering module is configured to: reorder each candidate information based on the information freshness of each candidate information in the candidate information queue and the corresponding expected revenue per thousand exposures, so as to obtain the first reordered candidate information of the candidate information queue; The recommendation module is further configured to: recommend the first rearrangement candidate information; The information freshness determination module includes: The information acquisition submodule has been shown, and it is configured to: acquire the information of the terminal device from the current moment. n The information already displayed in this session includes... n It is an integer greater than 0; The similarity matrix generation submodule is configured to: construct a similarity matrix based on the displayed information. n × n Similarity matrix S The similarity matrix S Each matrix element ,in α ij Indicates the first i The information already displayed is related to the first item. j The similarity of the displayed information is 0 < α ij ≤1, i and j All are integers, 0 < i ≤ n ,0< j ≤ n , and s ij =s ji ; The augmented similarity matrix generation submodule is configured to: traverse each candidate piece of information in the candidate information queue and construct an augmented similarity matrix corresponding to each candidate piece of information. ,in, S It is the similarity matrix, V It is the similarity vector of the candidate information relative to the already displayed information. V =[ v 1 , v 2 , ... , v n ],in v 1 , v 2 , ... , v n The aforementioned candidate information is respectively with n The similarity of the information already presented in the above clause, and v 1 , v 2 , ... , v n For each element in the matrix that is greater than 0 and less than or equal to 1, T represents the transpose of the matrix. The information freshness generation submodule is configured to: perform determinant calculation on the augmented similarity matrix of each candidate information in the candidate information queue, and use the values ​​obtained from the determinant calculation as the information freshness of each candidate information.

7. The information recommendation device as described in claim 6, wherein, The similarity between two pieces of information is calculated as equal to ,in, m An integer greater than or equal to 0 represents the direct, non-repeating distance between the two pieces of information.

8. The information recommendation device as described in claim 6, wherein, In the similarity matrix, when i ≠ j When, 0 < α ij <1.

9. The information recommendation device as described in claim 6, wherein, The reordering module includes: The recommendation metric generation submodule is configured to multiply the information freshness of each candidate information in the candidate information queue by its corresponding expected revenue per thousand exposures to obtain the corresponding recommendation metric. The first rearrangement candidate information determination submodule is configured to: select the candidate information with the highest recommendation index from among the candidate information in the candidate information queue as the first rearrangement candidate information.

Images